Air Quality Index Prediction Based On ARIMA And SVR Combined Model

With the rapid development of industry,the air quality in our country has been deteriorating.It is necessary to provide suggestions for people’s traveling and predict of the air quality accurately,which promise to pretreat the air pollution problem effectively and prevent the air pollution from further aggravating.In this paper,the air quality index,‘AQI data’,of Jinan from January 1,2018 to December 31,2018 and the data of six pollutants contained in Jinan AQI are collated as research data to study the features of Jinan AQI’s changing,by establishing a suitable time series model,ARIMA+SVR combination model,and making prediction the future AQI value of Jinan.Based on the descriptive statistical analysis on Jinan air quality index AQI sequence,including Analysis of Jinan AQI Status,frequency statistics and normal test,frequency statistics indicate that the main pollutants of AQI in Jinan,PM2.5 and PM10,have the same trend of increase and decrease.The proportion of days with good grade is the largest,and the proportion of days with qualified air quality is 64.1%.Mild pollution accounts(for)the majority of pollution days,indicating that there is a trend of improvement of the air quality in Jinan compared with previous years.The normal test indicates that Jinan AQI sequence does not obey the strict normal distribution and shows the characteristics of spike,thick tail and right deviation.Time series model and ARIMA+SVR combination model are established for Jinan air quality index AQI series.In this paper,ARIMA,ARMA(0,4)-GARCH(1,1)combined model and ARIMA+SVR combined model are established for Jinan AQI sequence.The research work is as follows:(1)The stability and pure randomness of Jinan AQI sequence are tested,and a single relative optimal model is established as ARIMA(1,0,0).(2)Observing the timing chart of Jinan AQI sequence and finding the possibility of heteroscedasticity,ARCH model is established and different distributions of residuals are tried.Finally ARMA(0,4)-GARCH(1,1)joint model under generalized error distribution(GED)is established to fit Jinan AQI sequence.(3)Considering that the time series model only considers time series factors and does not use other external information,it will have certain influence on the prediction accuracy of the model.Therefore,considering the prediction results of time series model as the input characteristic variable of SVR,ARIMA+SVR combination model is established to further predict and analyze Jinan AQI sequence,which is also the research feature of this paper.